Phunware’s mobile applications platform allows some of the biggest brands in the world to deliver great mobile experiences and reach mobile app users with relevant content. To achieve this at scale, Phunware uses Neo4j as the engine to power a knowledge graph connecting more than a billion nodes.
In this week’s five-minute interview (conducted at GraphConnect 2018 in NYC), we spoke with James Gray, Director of Product for Big Data at Phunware about their use of Neo4j .
What made you choose Neo4j?
James Gray: Our dataset is huge. We see over a billion mobile devices a month. We have over 1.5 billion nodes in our database and over 17 billion relationships. Performance at scale was critical for us, as was seeing query performance in milliseconds or seconds.
Another important factor in choosing Neo4j is the ability to show our customers what the data looks like. Graphs are very intuitive and Neo4j allows us to show how customers can use their data to derive value.
What have been the most surprising results you’ve seen with Neo4j?
Gray: The ability for people to see what their data looks like visually is striking. We’ve been able to uncover all kinds of data quality issues and see things that we can do to improve the data quality.
We’ve seen great performance. Processes that we used to run on our big data cluster took hours or days before we had Neo4j in place. We’re now doing those same routines and extracting data and knowledge within seconds. The pace at which we’re able to bring data together has changed dramatically.
Can you tell us about the Phunware Knowledge Graph?
Gray: We just came to market with Phunware Knowledge Graph. We’re super excited to enable customers to do many of the same things that we do internally, to allow them to build upon our dataset, extend it with all of the interesting data and domain knowledge that they have in their businesses and allow it to become a critical knowledge asset within their company.
Literally within minutes or seconds, customers generate mobile device audiences at scale given very complex queries. They target based upon the types of apps that people used, the types of locations that they’ve visited, the types of intent they have and their interests.
We use Neo4j inside our company, and then as a part of our solution, Neo4j is really the engine that powers our Phunware Knowledge Graph, and we’re super excited about that.
If you could go back to when you first started using graphs, what would you change?
Gray: I think we went about using graphs very methodically. We took some time to understand what the data was and to model it, describe it and connect it. When we brought all that data into the Neoj4 graph database physically, we knew it was going to give us the types of insights and the types of connections we wanted.
Maybe we would have looked at some of the data quality issues that were resident within the data. But frankly, a lot of those surfaced because we had them in the graph and we were able to uncover some anomalies in the data especially around some of the geolocation data. Again being able to get it inside the graph, we’re able to uncover things that we just couldn’t have if the data was in its raw form.
Want to share about your Neo4j project in a future 5-Minute Interview? Drop us a line at firstname.lastname@example.org
Read the White Paper